Javier Yu

I recently completed a Ph.D. in Aeronautics and Astronautics at Stanford University. I was advised by Professor Mac Schwager as part of the Multi-robot Sytems Lab and was funded on a NSF Graduate Research Fellowship.

As a roboticist I'm interested in practical solutions for problems at the intersection of learning and perception. My recent work is focused on how learned scene reconstructions can be used to safely and effectively improve robot performance.

LinkedIn  /  Google Scholar  /  Github

Email:

Research

SOUS VIDE is a novel simulator, training approach, and policy architecture for end-to-end visual drone navigation. Our trained policies exhibit zero-shot sim-to-real transfer with robust real-world performance using only on-board perception and computation.

SAFER-Splat (Simultaneous Action Filtering and Environment Reconstruction) is a real-time, scalable, and minimally invasive action filter, based on control barrier functions, for safe robotic navigation and teleoperation in a detailed map constructed at runtime using Gaussian Splatting.

Splat-Nav is a fast and safe robot navigation and localization pipeline designed to generate motion plans through Gaussian Splatting maps with rigorous collision constraints.

Nerf Bridge is an open-source and flexible software package that bridges between the ROS and the popular Nerfstudio library that enables real-time, online training of radiance fields (both NeRF and GS) using robot sensing data.

DiNNO is a distributed optimization algorithm that allows teams of robots to cooperatively optimize neural network models using fully decentralized communication and computation. With DiNNO robots do not share data with each other, but instead share model updates, allowing for privacy-preserving and efficient multi-robot learning.

Distributed optimization is provides an generalizable algorithmic framework for deriving distributed algorithms for multi-robot applications. In this two part journal series, we provide a tutorial for casting multi-robot problems as distributed optimization problems, and survey the state of the art in distributed optimization algorithms.

ICRA 2020 Best Multi-Robot Systems Paper Award

We present a distributed multi-target tracking algorithm that enables autonomous vehicle fleets to efficiently and collaboratively track hundreds of target vehicles.

A novel algorithm for online, distributed, non-myopic task selection for heterogeneous robot teams.

This website is adapted from Jon Barron's personal website.